Method and system for monitoring key equipment for ensuring power supply, electronic equipment and medium

Abstract

The invention provides an insurance power supply key equipment monitoring method and system, electronic equipment and a medium. The method comprises the following steps: collecting real-time state data of target insurance power supply key equipment based on a sensor; the target guaranteed power supply key equipment is any guaranteed power supply key equipment in the power system, and the real-time state data comprises operation state information, environment monitoring information and bearing load information at the current acquisition time t; controlling the target edge computing node to receive the real-time state data; inputting the operation state information, the environment monitoring information and the bearing load information into an equipment fault monitoring model to obtain equipment monitoring information output by the equipment fault monitoring model; the equipment fault monitoring model is obtained by training based on sample operation state information, sample environment monitoring information, sample bearing load information and corresponding alarm information and alarm category labels; and the operation state of the target guaranteed power supply key equipment is monitored based on the equipment monitoring information, so that the fault identification capability of the guaranteed power supply key equipment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment monitoring, and in particular to a method for monitoring key power supply equipment, a system thereof, electronic equipment and a medium. Background Art

[0002] The normal operation of key power supply equipment is an important device to ensure the safe and stable operation of the power system. Therefore, it is necessary to monitor in real time whether the key power supply equipment is in normal operation. For the monitoring of key power supply equipment, the existing method is mainly a rule-based fault diagnosis method, that is, the fault judgment rules are set according to expert experience. When the set fault judgment rules are met, it is determined that the monitored key power supply equipment is in an abnormal state. However, the rule-based fault diagnosis method relies on fixed rules and thresholds to determine whether the equipment is abnormal, which makes it unable to adapt to the complex and changeable equipment working environment, resulting in insufficient recognition of new types of faults, thereby reducing the fault recognition ability of key power supply equipment. Summary of the invention

[0003] The purpose of the present invention is to provide a key power supply equipment monitoring method and system, electronic device, and computer-readable storage medium to solve the technical problem that the existing equipment cannot adapt to the complex and changeable working environment, resulting in insufficient ability to identify new types of faults.

[0004] To achieve the purpose of the present invention, according to a first aspect of the present invention, a method for monitoring key power supply equipment is provided, comprising:

[0005] Collecting real-time status data of target power supply key equipment based on sensors; the target power supply key equipment is any power supply key equipment in the power system, and the real-time status data includes operating status information, environmental monitoring information and load information at the current collection time t;

[0006] Controlling the target edge computing node to receive the real-time status data;

[0007] The operation status information, the environmental monitoring information and the load bearing information are input into the equipment fault monitoring model to obtain the equipment monitoring information output by the equipment fault monitoring model; the equipment fault monitoring model is trained based on the sample operation status information, the sample environmental monitoring information, the sample load bearing information and their corresponding alarm information and alarm category labels;

[0008] The operating status of the target key power supply guarantee equipment is monitored based on the equipment monitoring information.

[0009] According to a second aspect of the present invention, a system for monitoring key power supply equipment is provided, characterized in that it includes a module for executing the method for monitoring key power supply equipment as described in the first aspect.

[0010] According to a third aspect of the present invention, there is provided an electronic device, comprising: a memory for storing a computer software program; a processor for reading and executing the computer software program, thereby implementing the method for monitoring key power supply equipment as described in the first aspect above.

[0011] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein a computer software program is stored in the storage medium, and when the computer software program is executed by a processor, the method for monitoring key power supply equipment as described in the first aspect above is implemented.

[0012] The method for monitoring key power supply equipment provided by the present invention monitors each key power supply equipment in the power system through an equipment fault monitoring model trained based on operating status information, environmental monitoring information, load information and their corresponding alarm information and alarm category labels, so that the key power supply equipment can be monitored in different working environments. Therefore, the types of failures of the key power supply equipment in different working environments can be identified, thereby improving the fault identification capability of the key power supply equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying creative labor, other drawings obtained based on these drawings still belong to the scope of the present invention.

[0014] Figure 1 It is a flow chart of a method for monitoring key power supply equipment in an embodiment of the present invention;

[0015] Figure 2 It is a structural diagram of a key power supply guarantee equipment monitoring system in an embodiment of the present invention;

[0016] Figure 3 is a diagram of an embodiment of an electronic device in an embodiment of the present invention;

[0017] Figure 4 Detailed description of the invention The figure is a diagram of an embodiment of a computer-readable storage medium in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The detailed description of the drawings is intended as an illustration of the current embodiment of the present application, and is not intended to represent the only form in which the present application can be implemented. It should be understood that the same or equivalent functions can be accomplished by different embodiments intended to be included in the spirit and scope of the present application.

[0019] Reference Figure 1 , Figure 1 : is a flow chart of a key power supply equipment monitoring method provided by an embodiment of the present invention. The execution subject of the key power supply equipment monitoring method provided by an embodiment of the present invention may be an equipment monitoring system. Therefore, the key power supply equipment monitoring method includes the following steps:

[0020] Step 10: Collect real-time status data of key power supply equipment based on sensors.

[0021] In the embodiment of the present invention, each key power supply guarantee device is installed with a corresponding sensor, so the status data of the key power supply guarantee device can be collected through the sensor.

[0022] Therefore, for any target power supply key equipment in the power system, its real-time status data at each time point can be collected through sensors on the target power supply key equipment.

[0023] In one embodiment, the sensor may include an operation status monitoring sensor, an environment monitoring sensor, and a load bearing monitoring sensor. Therefore, the collected real-time status data may include the operation status information, environment monitoring information, and load bearing information of the target key power supply equipment at each current collection time t. Among them, the operation status information in the embodiment of the present invention is the operation parameters, such as temperature, pressure, speed, current, etc. The environmental monitoring information is the surrounding environment data of the target key power supply equipment, such as humidity, vibration, noise level, external temperature, etc. The load bearing information is the working status of the target key power supply equipment under different loads, such as peak load, average load, load fluctuation data, etc.

[0024] Step 20, control the target edge computing node to receive real-time status data.

[0025] It should be noted that the sensor in the embodiment of the present invention is embedded with a corresponding communication module, which can transmit the collected real-time status data to the equipment monitoring system.

[0026] Therefore, in order to improve the data transmission rate and realize rapid monitoring of key power supply equipment, in an embodiment of the present invention, the equipment monitoring system communicates data with the sensor by calling the edge computing node in the data transmission network. In the data transmission network, the edge computing node is always in the state of processing the transmitted data. Since different edge computing nodes process different amounts of data, different edge computing nodes transmit data at different rates. Therefore, the equipment monitoring system needs to determine the optimal edge computing node (target edge computing node) in the data transmission network, that is, the edge computing node with the fastest data transmission rate, that is, the edge computing node that takes the shortest time for the equipment monitoring system to receive real-time status data, and control the target edge computing node to receive the real-time status data transmitted by the sensor.

[0027] Step 30, inputting the operation status information, environmental monitoring information and load bearing information into the equipment fault monitoring model to obtain the equipment monitoring information output by the equipment fault monitoring model.

[0028] Optionally, after receiving the operating status information, environmental monitoring information and load bearing information, the equipment monitoring system inputs the operating status information, environmental monitoring information and load bearing information into a pre-trained equipment fault monitoring model. The equipment fault monitoring model performs data processing and monitoring and prediction on the operating status information, environmental monitoring information and load bearing information, and outputs corresponding equipment monitoring information.

[0029] The equipment fault monitoring model in the embodiment of the present invention is obtained by training based on sample operation status information, sample environmental monitoring information, sample load bearing information and their corresponding alarm information and alarm category labels.

[0030] Step 40: monitor the operating status of the target key power supply guarantee equipment based on the equipment monitoring information.

[0031] Optionally, the equipment monitoring information output by the equipment failure monitoring model in the embodiment of the present invention includes equipment status information, equipment failure type and failure probability, wherein the equipment status information includes whether the target key power supply equipment is in normal operating state and in abnormal state, and the abnormal state is when the target key power supply equipment fails.

[0032] Furthermore, the equipment monitoring system monitors the operating status of the target key power supply equipment based on the equipment status information, equipment failure type and failure probability.

[0033] The embodiment of the present invention monitors each key power supply equipment in the power system through an equipment fault monitoring model trained based on operating status information, environmental monitoring information, load information and their corresponding alarm information and alarm category labels, so that the key power supply equipment can be monitored in different working environments. Therefore, the types of failures of the key power supply equipment in different working environments can be identified, thereby improving the fault identification capability of the key power supply equipment.

[0034] In one embodiment, step 20 controls the target edge computing node to receive real-time status data, including:

[0035] Step 201, determining the data processing time of each edge computing node based on the data processing capability, current processing data volume, and to-be-processed data volume of each edge computing node in the data transmission network;

[0036] Step 202, determining the data transmission delay time of each edge computing node based on the channel power gain and data transmission rate of each edge computing node;

[0037] Step 203, determining the data transmission consumption time of each edge computing node based on the data processing time and the data transmission delay time of each edge computing node;

[0038] Step 204: determine a target edge computing node in the data transmission network based on the data transmission consumption time of each edge computing node.

[0039] Optionally, the embodiment of the present invention takes into account that different edge computing nodes have different data transmission rates. Therefore, the device monitoring system obtains the data processing capability, current processing data volume, and pending data volume of each edge computing node in the data transmission network, and calculates the data processing time of each edge computing node according to the data processing capability, current processing data volume, and pending data volume of each edge computing node. The specific calculation formula for the data processing time is as follows:

[0040]

[0041] Among them, T d1 represents the data processing time, D a Indicates data processing capability, N c Indicates the current amount of data being processed, N f Represents the amount of data to be processed, ε1 and ε2 represent the preset weights. Generally, the preset weight ε1 is set to 0.6-0.8, and the preset weight ε2 is set to 0.3-0.5.

[0042] Furthermore, the device monitoring system obtains the channel power gain and data transmission rate of each edge computing node, and calculates the data transmission delay time of each edge computing node according to the channel power gain and data transmission rate of each edge computing node. The specific calculation formula of the data transmission delay time is as follows:

[0043]

[0044] Among them, T d2 represents the data transmission delay time, B represents the channel bandwidth of the edge computing node, K represents the channel capacity of the edge computing node, and H B represents the channel power gain, P B represents the data transmission rate, and σ represents the channel gain factor.

[0045] Furthermore, the device monitoring system calculates the data transmission consumption time of each edge computing node according to the data processing time and data transmission delay time of each edge computing node. The specific calculation formula for the data transmission consumption time is as follows:

[0046] T all =β1*T d1 +β2*T d2

[0047] Among them, T all represents the time consumed by data transmission, β1 and β2 represent the time consumption weights, where the sum of the time consumption weight β1 and the time consumption weight β2 is equal to 1. Generally, the time consumption weight β2 is greater than the time consumption weight β1, and the two are not the same.

[0048] Furthermore, the equipment monitoring system compares the data transmission consumption time of each edge computing node, and determines the edge computing node corresponding to the minimum data transmission consumption time as the target edge computing node in the data transmission network.

[0049] The embodiment of the present invention determines the target edge computing node with the shortest data transmission time in the transmission network through the data processing capability, current processing data volume, to-be-processed data volume, channel power gain and data transmission rate of each edge computing node, and receives real-time status data through the target edge computing node, thereby improving the data transmission rate and realizing rapid monitoring of key power supply equipment.

[0050] In one embodiment, the training step of the equipment fault monitoring model includes:

[0051] Step 50, collecting sample operation status information, sample environment monitoring information and sample load bearing information;

[0052] Step 60, generating a sample multi-dimensional data feature based on the sample collection time of the collected sample information, the sample operation status information, the sample environment monitoring information and the sample load bearing information;

[0053] Step 70, based on the sample multi-dimensional data features and the corresponding alarm information and alarm category labels, a preset neural network model is trained to obtain an equipment fault monitoring model.

[0054] Optionally, the equipment monitoring system collects sample operation status information, sample environment monitoring information and sample load bearing information.

[0055] Furthermore, the equipment monitoring system performs feature extraction on the sample operation status information, the sample environment monitoring information and the sample load bearing information, such as the process of step 301 of the embodiment.

[0056] Furthermore, the equipment monitoring system extracts the sample operation status information, the sample environment monitoring information and the sample load bearing information to obtain the sample data features, and performs dimension adjustment with the sample collection time of the sample information to generate the sample multi-dimensional data features, as shown in the specific process of step 302 and steps 3021 to 3024.

[0057] Furthermore, the equipment monitoring system inputs the sample multidimensional data features and the corresponding alarm information and alarm category labels into a preset neural network model, performs model training on the preset neural network model, and obtains an equipment fault monitoring model. The preset neural network model may be a feedforward neural network (FNN), a convolutional neural network (CNN), a recurrent neural network (RNN), a long short-term memory network (LSTM), etc.

[0058] It should be noted that the neural network model of the embodiment of the present invention is a long short-term memory network LSTM, that is, the equipment fault monitoring model of the embodiment of the present invention is obtained based on the training of the long short-term memory network LSTM.

[0059] Taking into account the prediction of the equipment fault monitoring model for different categories, as well as the category imbalance problem and prediction uncertainty problem, the total loss function Loss of the equipment fault monitoring model in the embodiment of the present invention includes the classification loss function L cls , weighted loss function L w and uncertainty loss function L uncertainty ;

[0060] Therefore, the total loss function Loss of the equipment fault monitoring model can be expressed as:

[0061] Loss = L cls +λ1L w +λ2L uncertainty

[0062] Among them, the classification loss function L cls It can be expressed as:

[0063]

[0064] Among them, the weighted loss function L w It can be expressed as:

[0065]

[0066] Among them, the uncertainty loss function L uncertainty It can be expressed as:

[0067]

[0068] Therefore, the final total loss function Loss of the equipment fault monitoring model can be expressed as:

[0069]

[0070] Among them, N represents the number of samples, C represents the number of categories, and y i,c represents the alarm category label of the cth category of the ith sample, represents the probability predicted by the model for the cth category of the ith sample, w i represents the weight for the i-th sample, which is dynamically set based on the sample running status information. represents the prediction variance for the i-th sample, λ1 represents the first preset hyperparameter, and λ2 represents the second preset hyperparameter.

[0071] The embodiment of the present invention trains an equipment fault monitoring model that can be applied to complex and changeable environments by collecting time, operating status information, environmental monitoring information, load information and their corresponding alarm information and alarm category labels, thereby improving the model's ability to identify new types of faults. Therefore, each key power supply equipment in the power system is monitored through the equipment fault monitoring model, so that the key power supply equipment can be monitored in different working environments. Therefore, the types of failures of key power supply equipment in different working environments can be identified, thereby improving the fault identification ability of key power supply equipment.

[0072] In one embodiment, the equipment fault monitoring model includes a feature extraction layer, a data dimension processing layer, an equipment state prediction layer, a fault type determination layer, and a fault occurrence probability prediction layer. Among them, the feature extraction layer is used to extract data features in real-time state data. The data dimension processing layer is used to splice the extracted data features to obtain multi-dimensional data features. The equipment state prediction layer is used to predict the equipment state of key power supply guarantee equipment. The fault type determination layer is used to predict the equipment fault type of key power supply guarantee equipment. The fault occurrence probability prediction layer is used to predict the fault occurrence probability of key power supply guarantee equipment.

[0073] Therefore, step 30 inputs the operation status information, environmental monitoring information and load bearing information into the equipment fault monitoring model to obtain the equipment monitoring information output by the equipment fault monitoring model, including:

[0074] Step 301, input the operation status information, environmental monitoring information and load bearing information into the feature extraction layer for feature extraction, and obtain the operation status features, environmental monitoring features and load bearing features output by the feature extraction layer.

[0075] The feature extraction layer in the embodiment of the present invention is a multi-layer channel convolution feature extraction layer, which is implemented based on the self-attention mechanism. In one embodiment, the feature extraction layer may include a convolution layer, an activation function ReLU and a pooling layer. The specific process is: the operating status information is input into the first channel convolution feature extraction layer of the feature extraction layer, and the operating status information is extracted through the first channel convolution feature extraction layer to obtain the operating status features output by the first channel convolution feature extraction layer. Similarly, the environmental monitoring information is input into the second channel convolution feature extraction layer for feature extraction, and the load information is input into the third channel convolution feature extraction layer for feature extraction to obtain the environmental monitoring features output by the second channel convolution feature extraction layer and the load characteristics output by the third channel convolution feature extraction layer.

[0076] Step 302: input the operating status characteristics, environmental monitoring characteristics and load bearing characteristics into the data dimension processing layer for dimension adjustment to obtain N*4 dimensional data characteristics output by the data dimension processing layer.

[0077] Optionally, the equipment monitoring system inputs the operating state characteristics, environmental monitoring characteristics and load characteristics into the data dimension processing layer for dimension adjustment, and obtains N*4-dimensional data characteristics output by the data dimension processing layer, wherein N is determined based on the dimension of the current acquisition time t. In one embodiment, the current acquisition time t includes t1, t2, t3 and t4, the current acquisition time t1 corresponds to the operating state characteristics a1, the environmental monitoring characteristics a2 and the load characteristics a3, the current acquisition time t2 corresponds to the operating state characteristics b1, the environmental monitoring characteristics b2 and the load characteristics b3, the current acquisition time t3 corresponds to the operating state characteristics c1, the environmental monitoring characteristics c2 and the load characteristics c3, and the current acquisition time t4 corresponds to the operating state characteristics d1, the environmental monitoring characteristics d2 and the load characteristics d3, therefore, the N*4-dimensional data characteristics can be expressed as {(t1*a1*a2*a3), (t2*b1*b2*b3), (t3*c1*c2*c3), (t4*d1*d2*d3)}.

[0078] Step 303: input the N*4 dimensional data features into the device state prediction layer to obtain the device state information output by the device state prediction layer.

[0079] Optionally, the equipment monitoring system inputs N*4 dimensional data features into the equipment status prediction layer to obtain equipment status information output by the equipment status prediction layer. The equipment status information includes whether the target key power supply equipment is in normal operating state and in abnormal state. The abnormal state means that the target key power supply equipment has a fault.

[0080] Step 304: input the N*4-dimensional data features into the fault type determination layer to obtain the equipment fault type output by the fault type determination layer.

[0081] Furthermore, the equipment monitoring system inputs the N*4-dimensional data features into the fault type determination layer to obtain the equipment fault type output by the fault type determination layer. The equipment fault types include mechanical faults, electrical faults, temperature anomalies, communication faults, software faults, environmental factors, etc. Among them, mechanical faults include bearing damage, gear wear or breakage, motor failure, etc. Electrical faults include short circuit or open circuit, insulation aging or breakdown, current overload, etc. Temperature anomalies include overheating (causing equipment protection shutdown), cooling system failure, etc. Communication faults include signal loss and data mistransmission between the control system and the equipment. Software faults include control system program crashes and parameter setting errors. Environmental factors include corrosion caused by excessive humidity, and dust or pollutants affecting equipment operation.

[0082] Step 304: input the N*4 dimensional data features into the fault occurrence probability prediction layer to obtain the fault occurrence probability output by the fault occurrence probability prediction layer.

[0083] Step 305: determine the device status information, the device failure type and the failure probability as the device monitoring information.

[0084] Furthermore, the equipment monitoring system inputs the N*4-dimensional data features into the fault occurrence probability prediction layer to obtain the fault occurrence probability output by the fault occurrence probability prediction layer.

[0085] Furthermore, the equipment monitoring system determines the equipment status information, equipment failure type and failure probability as equipment monitoring information. In one embodiment, the equipment monitoring information is {equipment status information: normal operation status; equipment failure type; failure probability}, indicating that the key equipment for power supply is normal and has a very small probability of abnormality. The equipment monitoring information is {equipment status information: normal operation status; equipment failure type: temperature overheating; failure probability: 20%}, indicating that the key equipment for power supply is normal, and there is a 20% probability that the key equipment for power supply will operate abnormally due to the abnormal temperature overheating. The equipment monitoring information is {equipment status information: abnormal status; equipment failure type: bearing damage; failure probability: 100%}, indicating that the key equipment for power supply has a mechanical failure of bearing damage.

[0086] The embodiment of the present invention monitors each key power supply equipment in the power system through an equipment fault monitoring model, so that the key power supply equipment can be monitored in different working environments. Therefore, the types of failures of the key power supply equipment in different working environments can be identified, thereby improving the fault identification capability of the key power supply equipment.

[0087] In one embodiment, the data dimension processing layer includes a first fully connected layer, a second fully connected layer, two residual convolution layers and a feedforward network layer. The first fully connected layer and the second fully connected layer are used to fully connect the data features. The two residual convolution layers are used to align and splice the data features after the fully connected layers are fully connected. The feedforward network layer is used to map the data features after the splicing of the two residual convolution layers with time.

[0088] Step 302 inputs the operating status characteristics, environmental monitoring characteristics and load characteristics into the data dimension processing layer for dimension adjustment to obtain N*4 dimensional data characteristics output by the data dimension processing layer, including:

[0089] Step 3021: Input the operating status features and the environmental monitoring features into the first fully connected layer, and fully connect the operating status features and the environmental monitoring features based on the first fully connected layer to obtain the first fully connected features output by the first fully connected layer.

[0090] Step 3022: Input the running status characteristics and the load bearing characteristics into the second fully connected layer, and fully connect the running status characteristics and the load bearing characteristics based on the second fully connected layer to obtain the second fully connected characteristics output by the second fully connected layer.

[0091] The fully connected layer in the embodiment of the present invention uses a convolutional layer and a pooling layer. In the process of fully connected processing, the data features are used as the input of the convolutional layer, and a bias is added to improve the generalization ability of the features. Then, the features are activated by a nonlinear activation function to perform full connection of the features to obtain fully connected features. Therefore, the fully connected process of the fully connected features is as follows: FullyConnected = f(w a *X n +b1).

[0092] Among them, Fully Connected represents the fully connected feature, f(·) represents the nonlinear activation function of the Rectified Linear Unit (ReLU), and w a Represents the convolution kernel weight matrix of the convolution layer, X n represents the data features of the input convolution layer, n is determined according to the number of data features of the input convolution layer, and b1 represents the bias vector of the nonlinear activation function.

[0093] Therefore, the equipment monitoring system inputs the operating status features and the environmental monitoring features into the first fully connected layer, and fully connects the operating status features and the environmental monitoring features based on the first fully connected layer to obtain the first fully connected features output by the first fully connected layer. Therefore, the full connection process of the first fully connected features is as follows:

[0094] Fully Connected1=f(w a *X s *X r +b1)

[0095] Among them, Fully Connected1 represents the first fully connected feature, X s Indicates the running status characteristics, X r Represents environmental monitoring characteristics.

[0096] Furthermore, the equipment monitoring system inputs the operating status characteristics and the load characteristics into the second fully connected layer, and performs full connection on the operating status characteristics and the load characteristics based on the second fully connected layer to obtain the second fully connected feature output by the second fully connected layer. Therefore, the full connection process of the second fully connected feature is as follows:

[0097] Fully Connected2=f(w a *X s *X f +b1)

[0098] Among them, Fully Connected2 represents the second fully connected feature, X s Indicates the running status characteristics, X f Represents environmental monitoring characteristics.

[0099] Step 3023: input the first fully connected feature and the second fully connected feature into the two-layer residual convolution layer, perform residual connection on the first fully connected feature and the second fully connected feature based on the two-layer residual convolution layer, and obtain the residual splicing feature output by the two-layer residual convolution layer.

[0100] The two-layer residual convolution layer in the embodiment of the present invention is composed of the sigmoid residual activation function and the tanh residual activation function. Therefore, in the process of residual connection, the output result of the sigmoid residual activation function and the output result of the tanh residual activation function are actually multiplied to obtain the residual splicing features output by the two-layer residual convolution layer.

[0101] Optionally, the equipment monitoring system inputs the first fully connected feature and the second fully connected feature into two residual convolution layers, performs residual connection on the first fully connected feature and the second fully connected feature through a sigmoid residual activation function to obtain a first residual connection feature, and performs residual connection on the first fully connected feature and the second fully connected feature through a tanh residual activation function to obtain a second residual connection feature.

[0102] The specific process of sigmoid residual activation function is as follows:

[0103] H t =σ t (W t ·[Fully Connected1,Fully Connected2]+b2)

[0104] The specific process of tanh residual activation function processing is as follows:

[0105] H c =tan(W c ·[Fully Connected1,Fully Connected2]+b3)

[0106] Among them, H t represents the first residual connection feature, W t represents the linear matrix of the sigmoid residual activation function, b2 represents the bias vector of the sigmoid residual activation function, σ t (·) represents the sigmoid residual activation function, H c represents the second residual connection feature, Wc represents the linear matrix of the tanh residual activation function, b3 represents the bias vector of the tanh residual activation function, and tan(·) represents the tanh residual activation function.

[0107] Furthermore, the equipment monitoring system connects the first residual feature H t and the second residual connection feature H c Multiply them together to obtain the residual concatenation feature H output by the two residual convolution layers.

[0108] Step 3024: Input the residual splicing features into the feedforward network layer. Based on the feedforward network layer, linear mapping is performed on the residual splicing features and the current acquisition time t to obtain the N*4 dimensional data features output by the feedforward network layer.

[0109] The feedforward network layer in the embodiment of the present invention is processed based on the sigmoid linear activation function. Therefore, the equipment monitoring system inputs the residual splicing features into the feedforward network layer, linearly maps the residual splicing features and the current acquisition time t based on the feedforward network layer, and obtains the N*4 dimensional data features output by the feedforward network layer. The processing process of the feedforward network layer in the embodiment of the present invention is:

[0110] F L =σ(W f ·[H,t]+b f )

[0111] Among them, F L represents N*4 dimensional data features, σ(·) represents the sigmoid linear activation function, W f The coefficient representing the linear relationship, b f Represents the bias of the linear relationship.

[0112] The following describes a key power supply guarantee equipment monitoring system provided by an embodiment of the present invention. The key power supply guarantee equipment monitoring system described below and the key power supply guarantee equipment monitoring method described above can refer to each other.

[0113] See also Figure 2 , Figure 2 The structural diagram of the key power supply equipment monitoring system provided by the embodiment of the present invention includes:

[0114] The acquisition module 210 is used to collect real-time status data of the target power supply key equipment based on sensors; the target power supply key equipment is any power supply key equipment in the power system, and the real-time status data includes operation status information, environmental monitoring information and load bearing information;

[0115] A receiving module 220 is used to control the target edge computing node to receive real-time status data;

[0116] Intelligent prediction module 230, used to input operation status information, environmental monitoring information and load bearing information into the equipment fault monitoring model, and obtain equipment monitoring information output by the equipment fault monitoring model; the equipment fault monitoring model is obtained by training based on sample operation status information, sample environmental monitoring information, sample load bearing information and their corresponding alarm information and alarm category labels;

[0117] The device status monitoring module 240 is used to monitor the operating status of the target key power supply guarantee equipment based on the device monitoring information.

[0118] The key power supply equipment monitoring system provided by the embodiment of the present invention monitors each key power supply equipment in the power system through an equipment fault monitoring model trained based on operating status information, environmental monitoring information, load information and its corresponding alarm information and alarm category labels, so that the key power supply equipment can be monitored in different working environments. Therefore, the types of failures of the key power supply equipment in different working environments can be identified, thereby improving the fault identification capability of the key power supply equipment.

[0119] See also Figure 3 , Figure 3 FIG. 1 is an embodiment diagram of an electronic device provided by an embodiment of the present invention. Figure 3 As shown, an embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, the following steps are implemented:

[0120] Collect real-time status data of target power supply key equipment based on sensors; the target power supply key equipment is any key power supply equipment in the power system, and the real-time status data includes operating status information, environmental monitoring information and load information at the current collection time t;

[0121] Control the target edge computing node to receive real-time status data;

[0122] Inputting the operation status information, environmental monitoring information and load bearing information into the equipment fault monitoring model, and obtaining the equipment monitoring information output by the equipment fault monitoring model; the equipment fault monitoring model is obtained by training based on the sample operation status information, sample environmental monitoring information, sample load bearing information and their corresponding alarm information and alarm category labels;

[0123] Monitor the operating status of key power supply equipment based on equipment monitoring information.

[0124] See also Figure 4 , Figure 4Detailed description of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. Figure 4 As shown, this embodiment provides a computer-readable storage medium 400, on which a computer program 311 is stored. When the computer program 311 is executed by a processor, the following steps are implemented:

[0125] Collect real-time status data of target power supply key equipment based on sensors; the target power supply key equipment is any key power supply equipment in the power system, and the real-time status data includes operating status information, environmental monitoring information and load information at the current collection time t;

[0126] Control the target edge computing node to receive real-time status data;

[0127] Inputting the operation status information, environmental monitoring information and load bearing information into the equipment fault monitoring model, and obtaining the equipment monitoring information output by the equipment fault monitoring model; the equipment fault monitoring model is obtained by training based on the sample operation status information, sample environmental monitoring information, sample load bearing information and their corresponding alarm information and alarm category labels;

[0128] Monitor the operating status of key power supply equipment based on equipment monitoring information.

[0129] On the other hand, an embodiment of the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored on a computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the power supply key equipment monitoring method provided by the above methods, the method includes:

[0130] Collect real-time status data of target power supply key equipment based on sensors; the target power supply key equipment is any key power supply equipment in the power system, and the real-time status data includes operating status information, environmental monitoring information and load information at the current collection time t;

[0131] Control the target edge computing node to receive real-time status data;

[0132] Inputting the operation status information, environmental monitoring information and load bearing information into the equipment fault monitoring model, and obtaining the equipment monitoring information output by the equipment fault monitoring model; the equipment fault monitoring model is obtained by training based on the sample operation status information, sample environmental monitoring information, sample load bearing information and their corresponding alarm information and alarm category labels;

[0133] Monitor the operating status of key power supply equipment based on equipment monitoring information.

[0134] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for monitoring key power supply equipment, characterized in that: include: Collect real-time status data of key power supply equipment based on sensors; The target key power supply guarantee equipment is any key power supply guarantee equipment in the power system, and the real-time status data includes the operation status information, environmental monitoring information and load bearing information at the current collection time t; Controlling the target edge computing node to receive the real-time status data; Inputting the operating status information, the environmental monitoring information and the load bearing information into an equipment fault monitoring model to obtain equipment monitoring information output by the equipment fault monitoring model; The equipment fault monitoring model is obtained by training based on sample operation status information, sample environmental monitoring information, sample load bearing information and their corresponding alarm information and alarm category labels; The operating status of the target key power supply guarantee equipment is monitored based on the equipment monitoring information.

2. The method for monitoring key power supply equipment according to claim 1, characterized in that: The equipment fault monitoring model includes a feature extraction layer, a data dimension processing layer, an equipment status prediction layer, a fault type determination layer and a fault occurrence probability prediction layer; The step of inputting the operating status information, the environmental monitoring information and the load bearing information into the equipment fault monitoring model to obtain the equipment monitoring information output by the equipment fault monitoring model includes: Inputting the operating state information, the environmental monitoring information and the load bearing information into the feature extraction layer for feature extraction, and obtaining the operating state features, environmental monitoring features and load bearing features output by the feature extraction layer; Input the operating state feature, the environmental monitoring feature and the load bearing feature into the data dimension processing layer for dimension adjustment, and obtain N*4 dimensional data features output by the data dimension processing layer; wherein N is determined based on the dimension of the current acquisition time t; Inputting the N*4 dimensional data features into the device state prediction layer to obtain device state information output by the device state prediction layer; Inputting the N*4-dimensional data features into the fault type determination layer to obtain the equipment fault type output by the fault type determination layer; Inputting the N*4 dimensional data features into the fault occurrence probability prediction layer to obtain the fault occurrence probability output by the fault occurrence probability prediction layer; The device status information, the device failure type and the failure occurrence probability are determined as the device monitoring information.

3. The method for monitoring key power supply equipment according to claim 2 is characterized in that: The data dimension processing layer includes a first fully connected layer, a second fully connected layer, two residual convolution layers and a feedforward network layer; the operating state characteristics, the environmental monitoring characteristics and the load characteristics are input into the data dimension processing layer for dimension adjustment to obtain N*4 dimensional data features output by the data dimension processing layer, including: Inputting the operating state feature and the environmental monitoring feature into the first fully connected layer, and performing full connection on the operating state feature and the environmental monitoring feature based on the first fully connected layer to obtain a first fully connected feature output by the first fully connected layer; Inputting the operating state feature and the load bearing feature into the second fully connected layer, and performing full connection on the operating state feature and the load bearing feature based on the second fully connected layer to obtain a second fully connected feature output by the second fully connected layer; Inputting the first fully connected feature and the second fully connected feature into the two-layer residual convolution layer, performing residual connection on the first fully connected feature and the second fully connected feature based on the two-layer residual convolution layer, and obtaining the residual splicing feature output by the two-layer residual convolution layer; The residual splicing feature is input into the feedforward network layer, and the residual splicing feature and the current acquisition time t are linearly mapped based on the feedforward network layer to obtain the N*4 dimensional data feature output by the feedforward network layer.

4. The method for monitoring key power supply equipment according to claim 1, characterized in that: The training steps of the equipment fault monitoring model include: Collect sample operation status information, sample environment monitoring information and sample load bearing information; Generate sample multidimensional data features based on the sample collection time of the collected sample information, the sample operation status information, the sample environment monitoring information and the sample load bearing information; Based on the sample multi-dimensional data features and the corresponding alarm information and alarm category labels, a preset neural network model is trained to obtain the equipment fault monitoring model.

5. The method for monitoring key power supply equipment according to claim 4 is characterized in that: The total loss function Loss of the equipment fault monitoring model includes the classification loss function L cls , weighted loss function L w And the uncertainty loss function L uncertainty ; The total loss function Loss can be expressed as: Loss=L cls +λ1L w +λ2L uncertainty ; The classification loss function L cls It can be expressed as: The weighted loss function L w It can be expressed as: The uncertainty loss function L uncertainty It can be expressed as: The final form of the total loss function Loss can be expressed as: Among them, N represents the number of samples, C represents the number of categories, and y i,c represents the alarm category label of the cth category of the ith sample, represents the probability predicted by the model for the cth category of the ith sample, w i represents the weight for the i-th sample, which is dynamically set based on the sample running status information. represents the prediction variance for the i-th sample, λ1 represents the first preset hyperparameter, and λ2 represents the second preset hyperparameter.

6. The method for monitoring key power supply equipment according to claim 1, characterized in that: The controlling target edge computing node receives the real-time status data, including: Determine the data processing time of each edge computing node based on the data processing capacity, current processing data volume, and pending data volume of each edge computing node in the data transmission network; Determine the data transmission delay time of each edge computing node based on the channel power gain and data transmission rate of each edge computing node; Determine the data transmission consumption time of each edge computing node based on the data processing time and data transmission delay time of each edge computing node; A target edge computing node in the data transmission network is determined based on the data transmission consumption time of each edge computing node.

7. The method for monitoring key power supply equipment according to claim 6, characterized in that: The calculation formula for determining the data processing time of each edge computing node is as follows: Among them, T d1 represents the data processing time, D a Indicates data processing capability, N c Indicates the current amount of data being processed, N f represents the amount of data to be processed, ε1 and ε2 represent the preset weights; The calculation formula for determining the data transmission delay time of each edge computing node is as follows: Among them, T d2 represents the data transmission delay time, B represents the channel bandwidth of the edge computing node, K represents the channel capacity of the edge computing node, and H B represents the channel power gain, P B represents the data transmission rate, σ represents the channel gain factor; The calculation formula for determining the data transmission consumption time of each edge computing node is: T all =β1*T d1 +β2*T d2 ; Among them, T all represents the time consumed for data transmission, and β1 and β2 represent the weights of the time consumption.

8. A key power supply equipment monitoring system, characterized in that: It includes a module for executing the method for monitoring key power supply equipment as described in any one of claims 1 to 7.

9. An electronic device, characterized in that: include: Memory for storing computer software programs; A processor is used to read and execute the computer software program, thereby implementing the method for monitoring key power supply equipment as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer software program, and when the computer software program is executed by the processor, the method for monitoring key power supply equipment as claimed in any one of claims 1 to 7 is implemented.

Images